Probability Based Power Aware Error Resilient Coding * Minyoung Kim, Hyuno Oh, Niil Dutt, Alex Nicolau, Nalini Venatasubramanian School of Information & Computer Science University of California, Irvine, CA 92697-42 {minyoun, hoh, dutt, nicolau, nalini}@ics.uci.edu Abstract Error resilient encoding in video communication is becoming increasingly important due to data transmission over unreliable channels. In this paper, we propose a new power-aware error resilient coding scheme based on networ error probability and user expectation in video communication using mobile handheld devices. By considering both image content and networ conditions, we can achieve a fast recoverable and energy-efficient error resilient coding scheme. More importantly, our approach allows system designers to evaluate various operating points in terms of error resilient level and energy consumption over a wide range of system operating conditions. We have implemented our scheme on an H.26 video codec algorithm, compared it with the previous AIR, GOP and PGOP coding schemes, and measured energy consumption and video quality on the IPAQ and Zaurus PDAs. Our experimental results show that our approach reduces energy consumption by 4%, 24% and 7% compared with AIR, GOP and PGOP schemes respectively, while incurring only a small fluctuation in the compressed frame size. In addition, our experimental results prove that our approach allows faster error recovery than the previous AIR, GOP and PGOP approaches. We believe our error resilient coding scheme is therefore eminently applicable for video communication on energy-constrained wireless mobile handheld devices.. Introduction * Recent advances in technology enable mobile handheld devices to be equipped with wireless interfaces and there will be growing demand for high quality mobile multimedia communications. However, wireless multimedia communications in the mobile handheld environment face several challenges, including high error rate, bandwidth variations, and limitations of the mobile devices such as battery lifetime constraints and the low CPU computation capability. To overcome the bandwidth limitation, there are several existing video coding techniques developed, for example, H.26 and MPEG, to compress raw video sequences to encoded bitstreams. These video encoding techniques exploit spatial and temporal correlation to achieve a high compression ratio, but they are usually unaware about the device status and the networ conditions during the coding process. Therefore, multimedia data encoding requires a large amount of * This wor was partially supported by NSF award ACI-2428. information, leading to high computation and communication energy consumption, and transmitting multimedia data over wireless networs can be very unreliable due to pacet loss. This problem should be solved with reasonable compression efficiency with high error resiliency considering resource constraints, which is a crucial factor for the real-time multimedia communication over error prone and lossy networ using mobile handheld devices. Video communication over unreliable networing environments is challenging since data loss and corruption from several reasons such as traffic congestion and physical channel failure affect video quality severely unless a guaranteed quality of service (QoS) is available between the source and the destination. Also, the spatio-temporal prediction encoding and variable length coding (VLC) of the source coding cause error propagation. Since spatio-temporal prediction requires the previous frame to reconstruct the current frame, a single error can lead to consecutive errors in the following frames. Liewise, because of VLC, a single bit error causes the decoder to lose a synchronization point that maes the following bits useless. Therefore, a variety of techniques have been proposed to enhance the resilience of the video data encoding against the pacet errors [, 2]. The most well recognized method is to insert intra-coding to mitigate the effect of error propagation in a predictive video compression algorithm [, 4,, 6]. However, inserting intracoding influences compression efficiency adversely since it tends to increase total length of the encoded bitstream. From this observation, the prior studies on error resilient video encoding mainly tried to find out a solution that maximizes bitstream robustness with low bit rate. Meanwhile, as mobile devices increasingly have video communication functionality, low power encoding has become important. Several encoding schemes have been proposed to reduce energy consumption for multimedia applications [9,,, 2]. However, these studies dealt with either error resilience or low power issues independently. We believe it is critical for both issues to be addressed together, especially in the context of energyconstrained mobile devices. In this paper, we propose a new energy-efficient, errorresilient encoding scheme. Especially, we note the dual role For every macro bloc in a predictive frame (P-frame), the encoder decides whether it already nows this bloc from the preceding frame or whether it's completely new. In the former case, it only encodes the differences (inter mode). In the latter case, it encodes the whole macro bloc (intra mode). Every macro bloc in an intra frame (I-frame) should be encoded in intra mode.
of intra-coding: not only does intra-coding improve error resilience, but it also contributes to reducing encoding energy consumption since it does not require motion estimation (which is the most power consuming operation in a predictive video compression algorithm). Indeed, the system designer will therefore need to evaluate the trade-offs between the error resiliency level, compression efficiency, and power consumption. In this paper, we focus our attention on these tradeoffs. Specifically, we (i) propose (Probability Based Power Aware Intra Refresh), a new energy-efficient and error-resilient encoding scheme, based on the networ condition and the image content, (ii) implement our scheme as well as other existing error resilient encoding schemes on an H.26 codec, (iii) extensively compare with other error resilient encoding schemes in the context of error resiliency vs. encoding efficiency (both bit rate and energy consumption), and (iv) evaluate the tradeoffs between the error resiliency level, compression efficiency, and energy consumption on top of a real implementation platform. Our performance results indicate that saves as much as 7% to 4% energy compared with other error resilient techniques allowing faster error recovery than the previous approaches. This paper is organized as follows. In the next section, we briefly review previous wors on error resilient coding scheme. In Section, we state the problem we are addressing and discuss in detail the proposed technique. Section 4 presents experimental setup and results. In Section we draw conclusions and comment on possible extensions of this wor. 2. Error Resilient Coding original video reconstructed video Video Encoder ME DCT Q VLC Video Decoder MC IDCT DeQ VLD Multiplexing, Pacetizing & Channel Encoding Demultiplexing, Depacetizing & Channel Decoding Lossy Networ Figure. A typical video communication system Figure shows a typical video communication system. The orignal video is first compressed by a video encoder and the encoded bit stream is then multiplexed and pacetized. After pacetization, the bit stream usually undergoes a channel coding stage unless the networ guarantees error free transmission. At the receiver end, the transmitted bit stream should be decoded to reconstruct the original video. Therefore, it is important to devise video encoding schemes that can mae the compressed bit stream resilient to transmission errors since, in practice, current networ environments only support lossy data transfer. Error control can be carried out at different levels from the application (codec) layer to the networ transport layer. A good survey on error resilient coding techniques for real-time video communication can be found in [, 2]. At the networ transport layer, channel coding such as forward error correction (FEC) and automatic repeat request (ARQ) are applicable. However, these methods cannot guarantee complete recovery from errors and the decoder may still experience erroneous data streams. To mae matters worse, these errors propagate throughout the subsequent frames since encoding is based on the difference between successive frames. To reduce these effects, the encoder should consider error-resilience and generate a more robust bitstream that will not be affected by transmission losses. The most intuitive way to produce a robust bitstream is to insert intra frames (I-frames) periodically. In this group of picture (GOP) structure, one GOP is treated as an independent decodable entity. In other words, an I-frame serves as a refresh which cleans up any errors that have been propagated in the video sequence. However, I-frames are usually much larger than predictively coded frames (Pframes). This leads to several transmission problems such as buffer overflow, higher delay and lin congestion due to periodic peas in the bit rate. Moreover, I-frames are much more sensitive to errors in the sense that loss of an I-frame significantly degrades the quality of the reconstructed image of the following P-frames. Techniques to overcome these problems are adaptive intra refresh (AIR) [, 6] and progressive group of picture (PGOP) [, 4]. AIR and PGOP scheme insert intra-coded macro blocs (MBs) to enhance the robustness of the encoded bitstream. AIR updates the specified number of MBs that have higher difference from the corresponding MBs in the previous frame while PGOP refreshes intra-coded MBs on a column-by-column basis from left to right. Both of them eliminate the need for I- frames which means the burden of refreshing is distributed throughout all the frames, thereby producing a much smoother output rate and enhanced robustness to errors. Nevertheless, these approaches focus only on enhancement of image quality ignoring resource constraints such as power consumption. However, resource constraints need to be managed effectively while ensuring the integrity of the image quality during video communication using mobile handheld devices. Indeed, since handheld devices (e.g. PDAs and cell phones) have limited power budget of a battery that directly affects the computational resources available for the application, there is a critical need to manage video quality within this resource budget. Therefore, in this paper we propose a new error resilient encoding technique, named (Probability Based Power Aware Intra Refresh) that can run at various operating points in accordance with resource constraints.. (Probability Based Power Aware Intra Refresh) The GOP scheme inserts an I-frame to refresh the video data while the AIR and PGOP schemes insert intra-coded MBs after the motion estimation (ME) process in Figure to alleviate the effect of error propagation in a predictive video compression algorithm. AIR inserts a pre-defined number of intra-coded MBs with the highest sum of absolute differences (SAD) or mean square error (MSE) value from the ME output. Even though PGOP inserts a pre-defined number of columns of intra-coded MBs, PGOP also uses the ME output to generate stride bac MBs 2 to enhance image quality. Note 2 In order to prevent errors that may propagate across the column being refreshed, PGOP proposes to augment the refresh process to trap these propagations by refreshing the affected MBs. They call this as stride bac.
that AIR puts emphasis on the content awareness since it encodes the most active part of the frame while PGOP mainly pays attention to the networ status since it adapts the number of columns to be encoded as intra MBs based on pacet loss rate (PLR). However, probability based power aware intra refresh (), which we describe in the next subsections, is integrated into the ME process to determine the motion vector (content awareness); we therefore eliminate the unnecessary ME process based on PLR (networ status awareness) and thereby reduce the encoding energy consumption... The Algorithm For quantitative analysis, we denote each macro bloc (MB), the probability of correctness of the corresponding MB, and the networ pacet loss rate as σ, and α, m i, j, i, j respectively. Consider a quarter common intermediate format (QCIF) image, a video conferencing format with each frame containing 44 lines and 76 pixels per line: this means 9x MBs mi, j ( i < 9, j < ) with 6x6 pixels in a QCIF frame. Hence, we introduce a 9x matrix C that contains the probability of correctness σ i, j of the corresponding MB m i, j in the -th QCIF video frame as follows. σ, σ, σ, σ, σ, σ, C = σ8, σ8, σ8, We also introduce a user-defined parameter Intra_Th that captures user expectation about the error resiliency level. A higher Intra_Th value indicates a higher user expectation about bitstream robustness. Figure 2 illustrates the flow of our algorithm. At the beginning, we start with an error free image frame. As time goes by, re-evaluates the probability of correctness of each macro bloc to decide encoding mode and to find best matching macro bloc from the previous frame. The encoding mode selection is done by comparison between probability of correctness of a MB and a given threshold value Intra_Th. A MB with lower probability of correctness than Intra_Th should be encoded as intra MB since the Intra_Th values can be considered as requested error resiliency level. In other words, we can sip motion estimation in this case. For a MB that is determined to be encoded as an inter-coded macro bloc, motion estimation based on our heuristic that considers both networ condition and image content is required. Thus, our approach is integrated into the encoding process in two ways: (i) encoding mode selection and (ii) motion estimation. Now we consider the status of networ that can be expressed as pacet loss rate (PLR) and user expectation (Intra_Th) in encoding mode selection before motion estimation (ME). The first observation here is that the image quality is guaranteed while satisfying a given constraint. However, the more important contribution of this wor is that provides various operating points in terms of image quality and resource constraints. Note that can operate in various manners according to a given set of Proceed with next MB Start from error free image frame i, j set σ i, j = = Choose encoding mode based on probability of correctness σ i, j < Intra _ Th? No Yes Encode as Intra MB Last MB? Yes Update C & go to next frame No No ME based on our approach that considers probability of correctness as well as SAD Does the ME result satisfy SAD threshold? Yes Encode as Inter MB Figure 2. error resilient coding constraints. If a user defines Intra_Th value that approaches zero (meaning a user puts emphasis on compression efficiency), operates as if there is no error resilience feature at all. On the other hand, if user defined Intra_Th value equals to one, generates all macro blocs as intra macro bloc. The higher Intra_Th value (by which a user expects higher image quality), the more intra macro blocs will be generated. Similarly, as pacet loss ratio (PLR) grows, more intra macro blocs should be generated to guarantee performance requirement specified by Intra_Th. We illustrate our contributions in more detail through our experimental results in Section 4. In the following subsections, we discuss our heuristic extensively. Firstly, we address our heuristic for encoding mode selection and motion estimation considering error probability of each macro-bloc as well as SAD based on a probabilistic model. Then, we will explain how to update the probability of correctness of the current frame based on that of a previous frame to re-evaluate robustness of the current encoded bitstream.... Encoding Mode Selection Let us start with the first issue: how to use probability of correctness in encoding mode selection. As described in Figure 2, we can simply encode a MB as an intra-coded MB if it has lower probability of correctness than a given threshold value Intra_Th. The MB with low probability of correctness indicates that it is vulnerable to error propagation since that particular MB has already experienced a sufficient amount of inter-coding up to that point. We do not even have to go through motion estimation in this case. Note that this early decision improves total performance in terms of encoding time and energy, since motion estimation is very computation intensive in video compression.
..2. ME Based on Probability of Correctness As we mentioned in Figure 2, not only eliminates unnecessary computation required by motion estimation (ME) with early decision based on a probabilistic model, but also considers the probability of correctness in motion vector selection. Once a MB is determined to be encoded as inter MB, then motion estimation is required. In general, the motion estimation that generates the motion vectors to determine best matching bloc between the previous and current frame is solely based on the sum of absolute differences (SAD). As a result, a macro bloc with the lowest SAD value is chosen as a reference macro bloc regardless of error probability caused by pacet loss during transmission. Damaged Macro-bloc w/ lowest SAD Error-free Macro-bloc w/ highest SAD 2 Previous Frame Current Frame Figure. Motivational example for error resilient ME Figure illustrates the basic idea of our motion vector selection as a motivational example. Assume that there are three candidates for a reference macro bloc as shown in Figure. MB () has the lowest SAD value and probability of correctness among the candidates while MB () has the highest SAD value and probability of correctness. If we do not consider the networ pacet loss, we will simply choose MB (), the candidate with the lowest SAD value. Now, assume that MB () is damaged during transmission. In that case, a motion vector based on only the SAD value will choose the damaged macro bloc as a reference bloc which means image quality for that macro bloc will be degraded. This conventional approach results in low image quality if there is an error in the previous frame. Therefore, we should consider the influence of the networ pacet loss in the ME process. To tae networ pacet loss into account, we revise the SAD formulation to subsume probability of correctness and image difference (SAD). We choose a motion vector with higher probability of correctness and lower SAD value between a current MB and a candidate MB. For more details on the formulation, please refer to []. To summarize our decision process, Figure 4 shows the pseudo code for encoding mode selection. The inequality ( (SADmv SAD _ Th) > SADself ) is used in P- frame encoding to evaluate the encoding efficiency before we actually encode the MB with generated motion vector. The term SAD means SAD value between the current MB and mv the reference MB corresponding to the motion vector MV while SAD means the deviation of the pixel value in the self current MB itself. If the difference between SAD and mv SAD is not sufficient, then inter encoded MB probably will self generate more bits than intra encoded MB. Therefore, in that case the MB should be encoded as intra MB. ENCODING_MODE_SELECTION (C, α, Intra _ Th) - if (σi, - j < Intra _ Th) then Encoding as INTRA macro bloc else { Select motion vector [] if ( (SADmv SAD _ Th) > SADself ) then Encoding as INTRA macro bloc } Figure 4. Pseudo code for encoding mode selection... Update Probability of Correctness In this section, we will discuss how to generate the correctness matrix of current frame ( C ) from that of the previous frame ( C - ) with a given networ pacet loss rate α and a motion vector. In case of inter macro bloc, the matrix for probability of correctness of -th frame C can be calculated by Formula (): - σ i, j = ( α) min(σ of related MBs) - - + α ( similarity factor btw m i, j and m i, j ) σ i, j The first part of Formula () explains the situation when the previous frame is transmitted without networ pacet loss in which case the probability for error-free transmission of previous frame ( α) should be multiplied by the minimum probability of correctness of related macro blocs. The remaining part of Formula () indicates the situation when the previous frame experiences erroneous transmission such as pacet lost or data corruption whose probability can be expressed by pacet loss rate (PLR) α. The similarity factor depends on which error concealment algorithm we use at the decoder. Even when an image sample or several blocs of a sample are missing due to transmission errors, the decoder can try to estimate them based on the surrounding received samples, by maing use of inherent correlation among spatially and temporally adjacent samples. Such techniques are nown as error concealment techniques [2]. For instance, if we use a simple copy scheme from the corresponding MB of previous frame, we can calculate the similarity factor from SAD value between macro bloc m and i,j m i,. Note that we j can easily adopt various error concealment schemes to our approach by modifying the similarity factor. For intra macro bloc, Formula () can be reduced to Equation (2) since this macro bloc will serve as a refresh. σ i, j = ( α) - - + α ( similarity factor btw m i, j and m i, j ) σ i, j.2. Extension for Power Awareness With proper interfacing mechanisms between the codec (encoder/decoder) and the networ, can be easily modified to adjust its operations based on the networ conditions and user expectation. Considering Equation () from Section.., the probability of correctness of the -th frame can be approximately computed by Equation () if () (2)
there is no similarity between the consecutive frames and every frame can be encoded as inter frame: - σ i, j ( α) = () According to Equation (), if PLR ( α ) increases and Intra_Th is fixed, σ i, j decreases faster. Therefore, the inserts more intra macro blocs, which will result in the degradation of the encoding efficiency in terms of bit rate with less encoding energy. However, more intra macro blocs can guarantee the same level of error resiliency even though the PLR becomes higher. Furthermore, adapting (in this case, decreasing) the Intra_Th by the amount of the PLR increase can generate similar number of intra macro blocs. Liewise, if PLR decreases, we can increase the Intra_Th to encode with similar number of intra macro blocs. Note that more intra macro bloc represents higher error resiliency, less energy consumption, and less encoding efficiency. This flexibility enables to have power awareness in the sense that it can adaptively change its operating points either to guarantee image quality within a given power constraint or to minimize power consumption with satisfying a given image quality constraint. Based on the feedbac information from the networ, can be extended to adjust Intra_Th parameter to maximize error resilient level within current residual energy constraint. Liewise, with a given image quality level, can be extended to minimize energy consumption by adapting parameters. 4. Implementation and Evaluation In this section, we evaluate the performance of through extensive experiments including power measurement on PDAs. First, we compare our approach with existing error resiliency techniques: PGOP, GOP, and AIR. We present two sets of experiments: (a) the effect of error resiliency with respect to energy consumption, and (b) the variation of image quality with respect to error resiliency. Detailed experimental setup and results are discussed in our technical report []. 4.. Implementation Platform We have implemented a on the H.26 encoder [7] using fixed-point arithmetic since the PDAs that we used do not have a floating point unit. We assume that a simple copy scheme is used for error concealment at the decoding side. However, as we mentioned earlier, we can easily adopt various error concealment schemes by modifying the similarity factor in Equations () and (2). Note that we use a uniform distribution of frame discard to generate the pacet loss pattern. For simplicity, but without loss of generality, we use the frame loss rate to denote the networ pacet loss rate. For data transfer, we use the real time protocol (RTP) [8] and the variable-size encoded output of each frame is contained by a single pacet as long as it does not exceed the maximum transfer unit (MTU). For power measurement, we removed the internal battery from the PDA to measure the power consumption []. All our measurements were made using a National Instruments PCI DAQ (data acquisition) board to sample the voltage drop across the resistor at samples/second. We also use two different PDAs to verify our technique. The first PDA is a HP ipaq H with an Intel 4 MHz XScale processor with 28 MB SDRAM, 48 MB Flash ROM and integrated wireless. In the case of H, we installed Familiar.7.2 [] with GPE environment as an operating system. The second PDA is a Sharp Zaurus SL-6 with an Intel 4 MHz XScale processor with 2 MB SDRAM, 64 MB Flash ROM, and external Belin Compactflash FD66 wireless card. Sharp SL-6 is powered by Linux and the and Java based embedded OS and the Qtopia environment [4]. Display size for both of them is 24x2. All subsequent energy graphs depict the active energy, i.e., the total energy minus the idle energy. The results obtained allow us to derive the energy costs for encoding executions. 4.2. Basic Comparison with Existing Error Resilient Coding Techniques In this section, we compare existing error resilient techniques with to show the performance of our wor. Comparison is done with GOP, AIR [, 6], and PGOP [, 4]. Figure (a) and (b) demonstrate the image quality on varying different parameters with several existing error resilient coding techniques, where PLR is assumed to be %. represents that we encode without considering any error resiliency. PGOP-N indicates PGOP with N-columns refresh. In other words, N-columns from left to right in a frame should be always encoded as intra MBs to enhance robustness of the bitstream. On the other hand, GOP-N represents I:P ratio I:N where N is the number of P-frames per a single I-frame and AIR-N represents AIR with N intra MBs with the highest SAD values. We use the pea signalto-noise ratio (PSNR) and number of bad pixels as the image quality metric. Section 4.4 discusses the image quality metric in more detail. We choose Intra_Th that gives similar compression ratio with,, and as shown in Figure (c). Figure clearly shows that can generate the same quality of compressed image with less encoding energy consumption since our scheme sips motion estimation more frequently. Even though PGOP also sips motion estimation for the specific MBs in the refreshing column, it still requires motion estimation for stride bac MBs. This overhead will be larger with a small number of column refresh. In case of GOP, the image quality and encoding energy consumption should be similar with PGOP. In this experiment, GOP always generates a slightly smaller bitstream than other schemes because GOP generates fewer intra MBs. Hence, if we can adjust GOP to generate similar encoded file size, then the image quality and encoding energy consumption will be similar with PGOP except the overhead of stride bac. Lastly, AIR consumes a similar amount of the encoding energy without any error resilient scheme since AIR decides the encoding mode after motion estimation. reduces energy consumption due to early decisions in MB mode selection and generates a robust and even bitstream against networ pacet loss based on the probabilistic model. Figure 6(a) illustrates PSNR variation according to the networ pacet loss represented as from e to e7. For comparison, we choose PGOP-, GOP-8, and AIR- since those schemes generate a similar size of encoded bitstream. It should be pointed out that recovers faster than
Average PSNR (PLR = %) Number of Bad Pixels (PLR %) 6 PSNR (db) 24 8 2 6 Number of Bad Pixels (M) 4 2 foreman aiyo garden foreman aiyo garden (a) (b) Encoded File Size Encoding Energy Consumption (ipaq) 2 2 Size (KBytes) 8 6 4 2 Encoding Energy (J) 2 foreman aiyo garden (c) Figure. Comparison between and existing techniques, where PLR is assumed to be %: (a) the average PSNR (b) the number of bad pixels as an image quality measure (c) the encoded file size (d) the encoding energy consumption using ipaq as a performance measure (image sources: FOREMAN.QCIF, AKIYO.QCIF, and GARDEN.QCIF, frames) foreman aiyo garden (d) PNSR Variation Frame Size Variation 6 24 8 2 6 7 9 7 9 2 2 2 27 29 7 9 4 4 4 47 49 PSNR (db) Pacet Loss ee2 ee4 e e6 e7 2 2 Frame Number 7 9 7 9 2 2 2 27 29 7 9 4 4 4 47 49 Frame Size (Bytes) Frame Number PGOP- GOP-8 AIR- (a) Figure 6. Comparison between and existing techniques, where PLR is assumed to be %: (a) PSNR variation (b) frame size variation (image source: FOREMAN.QCIF, frames) PGOP and AIR, because our scheme not only has content awareness from the similarity factor but also has networ awareness from the probabilistic model of networ error. Even though GOP sometimes recovers faster than our scheme, GOP cannot guarantee rapid recovery from errors since GOP inserts an I-frame to refresh erroneous transmission. Thus, when GOP loses an I-frame due to networ error, it fails to reconstruct N consecutive P-frames. The error e7 shows the instance of I-frame loss. Therefore, in the worst case, GOP is able to guarantee error recovery only after N frames. Moreover, Figure 6(b) shows another drawbac of GOP: GOP generates an uneven bitstream that is undesirable from a communication perspective. The fact that the encoded frame size generated by GOP fluctuates severely will cause transmission problems such as buffer PGOP- GOP-8 AIR- (b) overflow, higher delay and lin congestion due to periodic peas in bit rate. 4.. Error Resiliency vs. Energy Consumption There are trade-offs in the number of intra macro blocs that directly affect the compressed size with a given PLR and Intra_Th (recall that an increased number of intra MBs results in a larger compressed bitstream). The encoding results demonstrate that can generate an encoded bitstream with various error resiliency levels since inserting more intra MBs leads to a more robust bitstream. Also, considering Intra_Th is a user-defined parameter that reflects user expectation about the error resiliency level, it should be noted that our algorithm covers all possible error resiliency levels: From Intra_Th =, (which means a user wants to
encode whole frames as intra MBs for maximum error resilience) to Intra_Th = (indicating that a user wants to encode with maximum compression efficiency, without any error resilience scheme). Besides, PLR equals to zero means we can encode whole frames as P-frames. However, if the PLR approaches, we need more Intra MBs to guarantee required robustness. From this, we are able to observe the trade-off between error resilient level and encoding energy consumption. We can easily expect that encoding energy consumption will be inversely proportional to the number of intra MBs, since intra coding does not require motion estimation. However, a larger number of intra blocs will result in more transmission due to the larger encoded bitstream. 4.4. Error Resiliency vs. Image Quality We now present the variation in image quality with respect to error resiliency. We use the pea signal-to-noise ratio (PSNR) as a quality metric, which is an indication of the distortion. We also use the number of bad pixels as a quality metric to overcome the limitation of the average PSNR since some reconstructed images with different errors have the same PNSR value. A pixel with significant difference from the original pixel value -- generated by either networ error or dependency among MBs in inter frame encoding -- is considered as a bad pixel. The number of bad pixels is a better metric than PSNR to represent error resiliency since it counts the number of pixels which will degrade perceptive quality while PSNR depends on the reconstructed value of the bad pixels and PSNR can vary due to different encoding scheme regardless of the pacet errors. Image quality varies with respect to different PLR and Intra_Th. As we explained in the previous section, a higher Intra_Th represents that a user requests more robust bitstreams. Therefore, the encoded bitstreams with higher Intra_Th value introduce a high PSNR value and smaller number of bad pixels.. Conclusions and Future Wor In this paper, we proposed -- Probability Based Power Aware Intra Refresh -- a new error resilient coding scheme, which is based on networ error probability and user expectation in video communication using mobile handheld devices. By considering both image content and networ condition, we can achieve a fast recoverable and energyefficient error resilient coding scheme. More importantly, we provide various operating points in terms of error resilient level and energy consumption over a wide range of system operating conditions. Our experimental results show that our approach can achieve the same compression efficiency with faster recovery and reduced energy consumption by 4%, 24% and 7% compared with the previous AIR, GOP and PGOP schemes respectively. We believe our error resilient coding scheme is therefore eminently applicable for video communication on energy-constrained wireless mobile handheld devices. Trade-offs between the power consumption and the error resilient level open a wide design space for future research subjects. Our future wor will aim to design proper interfacing mechanisms between the codec and the networ, so that the codec can adjust its operations based on the networ conditions to maximize its resource usage. We also see a more effective and less computationally intensive video quality measure and networ pacet error model for more accurate similarity factor. Cooperation with error control channel coding can be another interesting research topic since is independent from any other encoder and/or decoder side control mechanisms (i.e. rate control, channel coding, etc.). Further optimization, however, is possible if these control mechanisms are taen into consideration. Cooperation with traditional low power techniques such as dynamic voltage scaling (DVS) and dynamic frequency scaling (DFS) to explore more energy gain is also applicable as future research. References [] Y. Wang, S. Wenger, J. Wen, A. Katsaggelos, Review of Error Resilient Coding Techniques for Real-Time Video Communication, IEEE Signal Processing Magazine, vol. 7, no. 4, pp.6-82, Jul. 2. [2] Y. Wang, Q. Zhu, Error Control and Concealment for Video Communication: a Review, IEEE, Vol. 86, pp. 974-997. May 998. [] L. Cheng, M. Zari, "An Adaptive Error Resilient Video Encoder," SPIE Visual Communications and Image Processing (VCIP ), Jul. 2. [4] L. Cheng, M. Zari, "Perceptual Quality Feedbac Based Progressive Frame-level Refreshing For Robust Video Communication," IEEE Wireless Communications and Networing Conference (WCNC'4), Mar. 24. [] ISO/IEC 4496-2, Information Technology - Generic Coding of Audio-visual Objects Part 2, Mar. 2. [6] S. Worrall, A. Sada, P. Sweeney, A. Kondoz, Motion Adaptive Error Resilient Encoding for MPEG-4, IEEE International Conference on Acoustics, Speech, and Signal Processing, Vol., pp. 89-92, May. 2. [7] Draft ITU-T Recommendation H.26, Video Coding for Low Bitrate Communication, Mar. 996. [8] H. Schulzrinne, S. Casner, "RTP: A Transport Protocol for REAL-Time Applications", Internet Engineering Tas Force, Internet Draft, Oct. 2, 99. [9] C. Tayor, S. Dey, D. Panigrahi, Energy/Latency/Image Quality Tradeoffs in Enabling Mobile Multimedia Communication, Software Radio: Technologies and Services 2, pp.-66 [] C. Im, S. Ha, An Energy Optimization Technique for Latency and Quality Constrained Video Applications, First Worshop on Embedded Systems for Real-Time Multimedia (ESTIMedia ), 2, pp.8-2. [] S. Hua, G. Qu, S. Bhattacharyya, Energy Reduction Techniques for Multimedia Applications with Tolerance to Deadline Misses, Design Automation Conference 2, pp.-6 [2] Y. Cao, H. Yasuura, Quality-Driven Design by Bitwidth Optimization for Video Applications, Asia and South Pacific Design Automation Conference 2. [] http://familiar.handhelds.org [4] http://www.myzaurus.com/dev [] M. Kim, H. Oh, N. Dutt, A. Nicoulau, N. Vaatasuburamanian, : Probability Based Power Aware Intra Refresh, A New Energy-efficient Errorresilient Coding Scheme, CECS Technical Report #-, Univ. of California, Irvine, Feb. 2.